Saved in:
Bibliographic Details
Main Authors: Yan, Sikuan, Dong, Sicheng, Wang, Haotong, Nie, Ercong, Liu, Yilun, Bi, Jinhe, Xu, Yingjie, Schwarzmann, Susanna, Trivisonno, Riccardo, Tresp, Volker, Ma, Yunpu
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2605.17065
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914574369816576
author Yan, Sikuan
Dong, Sicheng
Wang, Haotong
Nie, Ercong
Liu, Yilun
Bi, Jinhe
Xu, Yingjie
Schwarzmann, Susanna
Trivisonno, Riccardo
Tresp, Volker
Ma, Yunpu
author_facet Yan, Sikuan
Dong, Sicheng
Wang, Haotong
Nie, Ercong
Liu, Yilun
Bi, Jinhe
Xu, Yingjie
Schwarzmann, Susanna
Trivisonno, Riccardo
Tresp, Volker
Ma, Yunpu
contents Memory has become an increasingly important component of agentic systems, as these systems are expected to reason over long-term experience. However, prior work has largely focused on unimodal memory, leaving multimodal memory relatively underexplored despite its central role in real-world applications. Compared with unimodal settings, multimodal memory introduces additional challenges, including heterogeneous input integration, person-centric information alignment, and evidence aggregation across different granularities. We present PyraVid, a hierarchical multimodal memory framework inspired by Event Segmentation Theory from cognitive science. PyraVid organizes long videos into a coarse-to-fine pyramid structure, enabling structured memory access and effective evidence aggregation. It further supports structure-guided memory expansion with pruning, allowing the retrieval of related events with strong causal connectivity but low semantic similarity while reducing noise. Experiments on multiple long-video understanding benchmarks show that PyraVid consistently improves performance across datasets, model scales, and question types, highlighting the effectiveness of hierarchical multimodal memory for long-horizon reasoning.
format Preprint
id arxiv_https___arxiv_org_abs_2605_17065
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle PyraVid: Hierarchical Multimodal Memory for Long-Horizon Video Reasoning
Yan, Sikuan
Dong, Sicheng
Wang, Haotong
Nie, Ercong
Liu, Yilun
Bi, Jinhe
Xu, Yingjie
Schwarzmann, Susanna
Trivisonno, Riccardo
Tresp, Volker
Ma, Yunpu
Multiagent Systems
Memory has become an increasingly important component of agentic systems, as these systems are expected to reason over long-term experience. However, prior work has largely focused on unimodal memory, leaving multimodal memory relatively underexplored despite its central role in real-world applications. Compared with unimodal settings, multimodal memory introduces additional challenges, including heterogeneous input integration, person-centric information alignment, and evidence aggregation across different granularities. We present PyraVid, a hierarchical multimodal memory framework inspired by Event Segmentation Theory from cognitive science. PyraVid organizes long videos into a coarse-to-fine pyramid structure, enabling structured memory access and effective evidence aggregation. It further supports structure-guided memory expansion with pruning, allowing the retrieval of related events with strong causal connectivity but low semantic similarity while reducing noise. Experiments on multiple long-video understanding benchmarks show that PyraVid consistently improves performance across datasets, model scales, and question types, highlighting the effectiveness of hierarchical multimodal memory for long-horizon reasoning.
title PyraVid: Hierarchical Multimodal Memory for Long-Horizon Video Reasoning
topic Multiagent Systems
url https://arxiv.org/abs/2605.17065